MUESLI: a free numerical and graphical library

older news

2019-10-11: Added binary packages of the last version for Ubuntu and Fedora.
2019-10-02: Release of the 2.15.7 version: mfChol and msLU are now able to process a (1,1) dense matrix, i.e. a scalar. The internal library SuiteSparse has been upgraded to the 5.4.0 version; METIS is no longer used. Few improvements in the msSet routine. The NULL driver is fixed to support the same number of simultaneous opened windows as the X11 driver. New routine mfFun2, which is similar to mfFun (apply a function element-wise to mfArray) but using two arguments. Memory leaks suppressed in mfMin and mfMax.
2019-05-15: Upload of a new archive for the source, dated 2019-05-15, which fix a bug in pgline when using GCC-9.1. Added binary versions for Ubuntu, Fedora and Mac OS X.
2019-05-11: Release of the 2.15.5 version: search of zeros of function now uses the Dekker algorithm (as in Matlab) instead of the classical Newton one. Compilation modes becomes DEBUG/OPTIM instead of DEBUG/RELEASE. ODE and DAE solvers support SPD (Symmetric and Positive Definite) jacobian matrices, and have improved CIC (Consistent Initial Conditions) computation when using sparse and banded jacobian. mfMin and mfMax Muesli fonctions are now consistent when using NaN values: by default NaNs are discarded (as in Matlab), but these special IEEE values can be taken into account via the optional argument 'nanflag'. Many small fixes in FML and FGL.
2018-12-07: Release of the 2.15.2 version: small fixes to get a correct compilation and installation on MacOS X. Added binary for MacOS Mojave.
2018-12-03: Added binary versions for SimplyFortran under Windows (64-bit only).
2018-11-21: Small fix in the FGL sources (PDF driver).
2018-11-13: Added binary versions for recent Fedora distributions (RPM format). Muesli version is now 2.15.1.
2018-10-10: Release of the 2.15.0 version. This is an important upgrade which concerns mostly the graphic library FGL. First a PDF driver has been added, which allow the creation of PDF images without converting the EPS files; in the same time the PNG driver has been removed. Second, thanks to libXft, antialiased characters are now used in the figures on the screen, giving very readable strings even when their size is small. Besides, binary packages are now provided for some recent versions of the Ubuntu linux family.
2018-06-25: Release of the 2.12.11 version. mfGradient extended to 2-rank mfArray. New 3rd-order derivative and periodic boundary conditions for mfSpline. New optional argument 'append' for msSaveAscii, which can write an mfArray at the end of an existing file. Double integration implemented (mfDblQuad). mfSimpson routine improved: now, the abscissa vector doesn't need to be equally spaced. Lastly, Finite-difference jacobian generation fixed for mfLsqNonLin.
2018-05-18: Release of the 2.12.10 version. New '.and.' and '.but.' operators for integer sequences used in mfGet and msSet. New mfPPDer routine which returns the derivative of a piecewise polynomial function.
2018-04-06: Release of the 2.12.9 version. mfInterp routines improved (both 1D and 2D interpolation). New derived type 'mf_Int_List' for manipulating and displaying integer lists. Updated 3D Delaunay tools (4 new routines).
2017-10-31: Release of the 2.12.8 version. New routines msBar and msErrorBar implemented. Configure script fixed for Ubuntu-17.10, and Fedora 24, 25, 26. About FileIO, new routine 'msLoadSparse()' added in order to get a sparse matrix and its right-hand-side vector from an HBO file. Improved detection of NaN values in some routines of the 'Optimization and Function functions' module. Fix the main Makefile for building the whole library under MacOS.
2017-08-29: Release of the 2.12.7 version. Few internal Muesli variables was not declared as private in the FML module, which prevented the user to use simple names: fixed. Improvements in the configure script, which better handles ZLIB and LIBPNG libraries. Redesign of the FSolve routine, which supports now sparse jacobian matrices. More features in the routine SpImport. Many bugs fixed in FML and FGL.
2017-06-29: New Binary version is provided for MinGW under Windows 64bit (we recall that FGL is partially available under Windows, i.e. only the non-interactive routines). See the Windows page.
2017-05-22: Release of the 2.12.4 version. Modification of the arg names of 'mfATan2' to be more clear: (X,Y) instead od (A,B). Any unwanted arg in 'mfOut' can be replaced by the keyword MF_NO_ARG. 'msReshape' routine (in-place version) now works for sparse matrices. Update ARPACK-NG library from 3.1.3 to 3.5.0. When using the ODE/DAE solvers, the user can name by groups not only his equations, but also his variables.
2017-05-06: Release of the 2.12.3 version. New FML routine 'msSmooth'. Added order 3 interpolation, both in 1D or 2D ('mfInterp1' and 'mfInterp2'). Messages on terminal are now colored. Few improvements or bugs fixed here and there. 'meditor' graphics tools has been ported from Qt3 to Qt4, and 'msEdit' works again.
2017-03-22: Release of the 2.12.1 version. Fixed or improved many routines of FML (LsqNonLin, Ode/DaeSolve) and FGL (msLegend, Pcolor, ImRead, ColorBar). In FGL, the color palette 'krahina' has been renamed 'fusion'.
2016-12-15: Release of the 2.11.3 version. Improvements of many routines, both in FML (mfGet, msSet, msFormat, mfLoadAscii, mfInterp1, mf_DE_Options, PDF documentation) and FGL (msPatch; few bugs fixed). Added Delaunay 3D (i.e. tetrahedra) set of routines. New mfExpInt routine (exponential integral).
2016-06-14: Release of the 2.11.1 version. New mfFun routine
2016-05-15: Release of the 2.11.0 version. Support of Darwin (Mac OS X, Yosemite) for both the numerical and the graphical part.
2016-04-29: Release of the 2.10.3 version. Few bugs fixed in the FML part. Automatic quadrature routine (mf/msQuad) improved. PNG driver improved for the creation of bitmap figures. Binaries versions are also provided for MinGW under Windows 64bit; for the first time FGL is also available under Windows but only the non-interactive routines are available.
2015-12-22: Release of the 2.10.2 version. Improvement in the FGL part (EPS driver) and few bugs fixed in the FML part. SuiteSparse package upgraded to 4.4.6.
2015-11-03: Release of the 2.10.1 version. Few improvements in the FGL part (background of figures is now white by default, not black; bold fonts are now supported both in X11 and EPS). Bug fix in msRat for very big real numbers.
2015-07-21: Release of the 2.10.0 version. Few bugs fixed in the graphics part (FGL). Added quadratic interpolation in mfInterp1. Fix bug for some FML routines when calling them with a temporary mfArray.
2015-06-16: Release of the 2.9.12 version. Many bugs fixed in the graphics part (FGL).
2015-04-24: Fix of the 2.9.11 version: processing of duplicated elements in sparse matrices has been fixed, so that the COO to CSC sparse conversion is now much faster: O(N) instead of O(N²).
2015-03-20: Release of the 2.9.11 version. Improved few routines (2nd-order schemes added in mfInterp2 and msGradient); many improvements in LsqNonLin; few bugs fixed.
2015-03-07: Release of the 2.9.10 version. Improved jacobian check for ODE, DAE solvers and LsqNonLin routine. Improved mfNull and mfRref routines. Few bugs fixed.
2015-03-03: Fix of the 2.9.9 version. Added stable box-constrained algorithm for LsqNonLin.
2015-02-25: Release of the 2.9.9 version. Improvements of the non linear solvers and integrators. Fix the automatic checks (quick and full) of the user-provided jacobian routine. Added the 'rational basis' optional case for the Reduced Row Echelon Form (msRref) routine.
2015-02-11: Release of the 2.9.8 version. Many improvements for the following routines: OdeSolve/DaeSolve, LsqNonLin. Due to the use of procedure pointers in LsqNonLin, the GCC-4.5 is the minimum version to use; therefore the GCC-4.4 support has been stopped.
2015-01-30: Updated binaries of Muesli-FML for use with Simply Fortran (Approximatrix) under Windows are available here.
2015-01-28: Release of the 2.9.7 version: Improvements of ODE/DAE solvers. Bug fixes.
2014-10-03: New 2.9.5 version: fix of the jacobian saving for DAE solver.
2014-08-18: Release of the 2.9.5 version. Improvements of the stopping conditions for the DaeSolve routine. Pay attention that the "End Condition" in OdeSolve/BDF is broken! Use DaeSolve/BDF instead... A future version should repair that...
2014-07-31: Release of the 2.9.4 version. Improvements of the OdeSolve routine: implementation of three different stopping conditions. Be aware that the API of the argument routines (deriv, jac, resid, ...) has been changed!
2014-06-23: Release of the 2.9.3 version, which fixes an important bug in DaeSolve routines.
2014-05-28: New archive of the 2.9.2 version (fix for an error in the test of mf/msLsqNonLin).
2014-05-01: Release of the 2.9.2 version. Many graphic improvements. SortRows is a new routine. LsqNonLin accepts now more optional arguments.
2014-03-14: Release of the 2.9.1 version. Improvements in some graphical routines (msText, mfGinput - be aware that the API of this latter routine has been changed). Bug fixes in read/write XPM images.
2014-02-21: Small fix for improved configure and Makefiles.
2014-02-18: Release of the 2.9.0 version. 'show_backtrace' is available for GCC version greater than 4.4 ; therefore, GCC-4.3 support has been stopped.
2014-02-05: Release of the 2.8.6 version. Few bugs fixed, especially in 'msPColor' (print in EPS is good now). 'msContour' has been improved, (labelling contours works now better) but its API has been modified. New routine 'mfCheckPerm' can be used to check whether a permutation vector is valid.
2013-10-27: Release of the 2.8.5 version. mf/msEigs now accept the following options for the eigenvalues selection: 'LA', 'SA', 'BE', 'LR', SR', 'LI, 'SI'. Analogously, mf/msSVDs accept 'BE'. The muesli_trace() becomes msMuesliTrace(). Fix important bug in mfDiag() about the 'symm' tag for complex matrices.
2013-10-09: Release of the 2.8.4 version. Muesli uses now arpack-ng the `new generation' version of Arpack (few eigenvalues/vectors of a big matrix), instead of the old original package (1996, patched 2002); it gives a more stable behavior, especially when using Simply Fortran (Approximatrix) under Windows. Fix also of a long waiting bug in the 'muesli-config' script for ubuntu-13.04: linking with the shared version of zlib cannot build successfully any executable that use only FML (and not FGL).
2013-10-02: Release of the 2.8.3 version. Improvements in the configure script. Few bugs fixed in the FGL part. 'msSaveAscii()' now writes numbers having an exponent larger than 100 under a standard way, i.e. with the "E" letter in front of the exponent. OPEN64 free fortran 90 compiler is no longer supported.
2013-08-20: Release of the 2.8.2 version. Random numbers are now generated via the RngStreams library, so that the generated numbers should be the same whatever the compiler used. Light modification of the API of the ODE/DAE solvers (t_span is a column vector; y_0 is a row vector). The DAE solver is now able to stop with two new conditions (Emergency Stop and End Condition), both must be set in the 'resid' routine (see the Reference Manual). SuiteSparse updated to version 4.2.1.
2013-07-07: Updated binaries of Muesli-FML for use with Simply Fortran (Approximatrix) under Windows are available here.
2013-07-05: Release of the 2.8.0 version: fix bug about the update of the internal structure properties of the mfArrays (TRIL,TRIU,SYMM,POSD) during operations on matrices. Better configure script for recent ubuntu distros. Muesli Binary File format (MBF) update to 2.4, in order to store all the matrix property inside the file (compatibility with 2.3 is ok); Matlab mex files have been updated accordingly. NONE support in image read/write (NONE means no color, i.e. transparency).
2013-02-03: Release of the 2.7.0 version: all the solvers of SuiteSparse use now LONG integers for the indexes. Added two new routines: 'mfExpm1()' and 'mfLog1p()' which are useful for sparse matrices.
2013-01-24: Binaries of few versions of MUESLI are available (including the last one, 2.6.6), for programmers using Simply Fortran under Windows. See the Windows page.
2013-01-18: For the windows users, discontinue to propose the MUESLI sources ready-to-compile for G95. Moreover, work is in progress to add some stuff about MinGW-w64 (a pretty good replacement for MinGW 32bit) and Simply Fortran (Approximatrix). See the Windows page.
Lastly, since early 2013 the IBM-xlf compiler is no longer supported. However, we continue to propose the 'configure' tool for building MUESLI.
2012-08-29: Fix regression which prevents to use the routines OdeSolve and DaeSolve.
2012-08-03: Release of the 2.6.6 version: thanks to Paweł Biernat, MUESLI can now be compiled on Linux distributions using PNG 1.5 (like ArchLinux). OdeSolve may now return (as DaeSolve) the whole history of the time steps used during the integration. Fixed configure script for recent Ubuntu 32bit versions.
2012-06-11: Release of the 2.6.5 version: fixed Makefiles for INTEL_IFC and OPEN64.
2012-05-08: Release of the 2.6.4 version: fixed a typo in the 'src/tests/fgl/run_all' script; increased the length of some character variables from 128 to 512.
2012-05-07: Update the web link to the Open64 free compiler.
2012-05-03: Release of the 2.6.3 version: the configure script has been polished for ubuntu-12.04. Two new routines have been added: mf/msFunFit() which make non-linear regression (non-linear fitting), mfInt() which converts a integer scalar mfArray into a f90 integer. The letter 'E' (for exponent) is now always used when writing floating-point reals in files (sparse matrices saved under the MTX format); before reals with a large exponent was written as -9.290690405090165-218 for example, which gave an error while reading with Matlab. Last, the MUESLI Installation Guide now emphasizes that a blank is not permitted in the full path of the directory containing the MUESLI sources.
2012-03-12: Release of the 2.6.2 version: two bug fixes in the previous release. First, a very annoying bug breaks the DAEsolver in DEBUG mode. Second, the Matlab mex files works now well for both 32-bit and 64-bit OS.
2012-03-07: Release of the 2.6.1 version: performance for ODE and DAE solvers (dense jacobian matrix) have been increased, due to the use of Lapack routines instead of old Linpack ones. Up to 50 % CPU time reduction can be obtained for large size differential systems.
2012-02-27: Release of the 2.6.0 version: the permutations used in many factorizations (LU, QR, CHOL, ...) are now returned under an integer vector form, embedded in a mfArray: this is the new data type MF_DT_PERM_VEC. The new routines mf/msColPerm, mf/msRowPerm, mf/msColScale and mf/msRowScale are used to apply a permutation or a scaling to a matrix. Upgrade to SuiteSparse-3.7.0. New global string variable: MF_LAPACK_LIB which contains the type of Lapack library used (Reference Blas/Lapack, MKL, Sunperf, ...).
2011-11-29: Release of the 2.5.2 version: improvement of the 'msDisplay()' routine, especially for sparse matrices. When solving a linear system which have no solution, RDiv and LDiv now return a vector of NaNs (instead of a vector of zeros). Improvement of ODE and DAE solvers: a structure change of the jacobian during the integration is now allowed. Small bug fixes in the sparse QR factorization. Improvement of the main 'configure' script for the MUESLI library. Upgrade to SuiteSparse-3.6.1.
2011-06-26: Release of the 2.5.1 version: new constant parameter MF_END, which is en automatic end-index for using in mfGet and msSet.
2011-06-19: Release of the 2.5.0 version: Important bug fixed concerning the computation of consistent initial conditions in the DAESolve routines. SuiteSparse SPQR has been updated to version 1.2.2. MUESLI now defines a new global variable, named MF_COMPILER_VERSION.
2011-05-31: Release of the 2.4.1 version: Many bug fixes in the graphics part (colorbar is printed correctly in EPS files; colorbar is well visible in PNG files; ...). Error and warning message's behaviour has been improved. 'mfColon' behaviour is also improved.
2011-02-12: Release of the 2.4.0 version: whole redesign of the Floating-Point Exceptions trapping, which are now dynamically treated (i.e. at run-time); this is a major improvement for debugging purposes.
2011-01-28: Release of the 2.3.2 version: upgrade of SuiteSparse (3.5.0); Lapack-3.3.0 tested; msLDLT now calls dsytrf of Lapack; Solaris Studio 12 support added; msMin and msMax doesn't return NaN any more for a sparse matrix.
2010-10-08: Release of the 2.3.0 version: major redesign of the printing process in EPS format (MUESLI now uses PostScript fonts). Few improvements and bug fixes in the DaeSolve family routines. API of mfSvd modified: it returns now a column vector instead of a diagonal matrix. mfAiry (Airy) and mfKron (Kronecker tensor product) functions added. mfReshape now works for sparse matrices.
2010-07-01: Release of the 2.2.5 version: Many bug fixes in the graphics part.
2010-05-18: Release of the 2.2.3 version: Many bug fixes (here and there) and improvements in the makefile options. Triangular and diagonal matrices are processed faster in mfLDiv and mfRDiv routines (API of these routines has been slightly modified — see documentation). Improvements in SVDS and Eigs when matrix is symmetric positive definite. Few routines renamed (mfHasDiagPattern, mfHasTrilPattern and mfHasTriuPattern become respect. mfIsDiag, mfIsTril and mfIsTriu). Improvements for output arguments in routine msDaeSolve (see documentation).
2010-02-20: Release of the 2.2.1 version: Bug fixes in DaeSolve family routine. Added support for the OPEN64 compilers.
2010-02-12: Release of the 2.2.0 version: MUESLI comes now with a configure script which attempts to check your system and compilers.
2010-01-29: Makefile fixed for FSF-g95.
2010-01-27: Few bugs fixed. Added a new archive for MinGW-gfortran.
2010-01-22: update the two packages. (in the mingw-g95 package, the 'GFT' directory was missing under 'misc').
2010-01-21: release of the 2.1.1 version, which introduces the 'mfBlkDiag()' routine for building a sparse matrix from square sparse blocks; as usual, few bugs fixed. Warning: the default behavior of the 'mfSpImport' routine has been changed for duplicated entries (see the Reference Manual).
2009-11-26: release of the 2.1.0 version: FFT and InvFFT have been added, both for vectors and matrices. FFT2 and InvFFT2 do two dimensional FFT on matrices.
2009-11-18: few bugs fixed, especially in the GNU_GFC Makefile, for the x86_64 architecture.
2009-11-13: the mingw version has been updated to the 2.0.4 version.
2009-10-15: release of the 2.0.4 version; many bugs fixed. The IBM/PowerPC platform is now supported (experimentaly, tested only on a Power6 machine with XLF 12.1 and XLC 10.1). SuiteSparse has been upgraded to 3.4.0. Few memory leaks fixed under FGL (thanks to the valgrind tool).
2009-09-11: release of the 2.0.3 version; many bugs fixed. Introduce also a work-around to use the 4.4/4.5 version of GCC.
2009-04-20: LAPACK 3.2.1 has been released. This is now the preferred version to be used for linking with MUESLI.
2009-02-02: release of the 2.0.1 version. HDF5 library is now optional, so this avoids many troubles when installing the F90 interface of HDF5. Besides, the Cholesky factorisation for sparse matrices may be splitted now is two steps: the symbolic one and the numeric one; this gives better performance when solving systems in loop, where only the numerical entries change.
2009-02-02: more clean in the Makefiles. New optional arguments for the mf/msContour routine.
2008-12-10: many bugs fixed. Makefiles have been cleaned. Three new routines in the 'elmat' module for inquiry about the pattern of a matrix (diag, tril, triu). Besides, the 'mf/msSchur' routine has a new optional argument: 'form'.
2008-11-28: fix for a missing lib in the Makefile for the FGL tests.
2008-11-27: fix the MinGW archive (2008-11-26): the main Makefile for FSF-g95 was missing. The Installation Guide also has been updated. Fix also a typo in the Makefiles for FSF-g95.
2008-11-26: release of the 2.0.0 version. No more features vs 1.9.9 but some bugs fixed.
2008-06-04: release of the 1.9.9 version.
2008-03-14: release of the 1.8.0 version.
2007-11-29: release of the 1.7.0 version.
2007-04-28: release of the 1.3.0 version.
2006-10-25: release of the 1.0.0 version.
2006-03-22: release of the 0.9.0 version.

Come back to the MUESLI home page.